Advances in the understanding of haemoglobin switching


  • Vijay G. Sankaran,

    1. Division of Hematology/Oncology, Children’s Hospital Boston, Harvard Medical School
    2. Department of Pediatric Oncology, Dana-Farber Cancer Institute, Harvard Stem Cell Institute, Harvard Medical School
    Search for more papers by this author
    • Contributed equally.

  • Jian Xu,

    1. Division of Hematology/Oncology, Children’s Hospital Boston, Harvard Medical School
    2. Department of Pediatric Oncology, Dana-Farber Cancer Institute, Harvard Stem Cell Institute, Harvard Medical School
    3. Howard Hughes Medical Institute, Boston, MA, USA
    Search for more papers by this author
    • Contributed equally.

  • Stuart H. Orkin

    1. Division of Hematology/Oncology, Children’s Hospital Boston, Harvard Medical School
    2. Department of Pediatric Oncology, Dana-Farber Cancer Institute, Harvard Stem Cell Institute, Harvard Medical School
    3. Howard Hughes Medical Institute, Boston, MA, USA
    Search for more papers by this author

Dr Stuart H. Orkin, Department of Pediatric Oncology, Dana-Farber Cancer Institute, 44 Binney St, Boston, MA 02115, USA. E-mail:


The study of haemoglobin switching has represented a focus in haematology due in large part to the clinical relevance of the fetal to adult haemoglobin switch for developing targeted approaches to ameliorate the severity of the β-haemoglobinopathies. Additionally, the process by which this switch occurs represents an important paradigm for developmental gene regulation. In this review, we provide an overview of both the embryonic primitive to definitive switch in haemoglobin expression, as well as the fetal to adult switch that is unique to humans and old world monkeys. We discuss the nature of these switches and models of their regulation. The factors that have been suggested to regulate this process are then discussed. With the increased understanding and discovery of molecular regulators of haemoglobin switching, such as BCL11A, new avenues of research may lead ultimately to novel therapeutic, mechanism-based approaches to fetal haemoglobin reactivation in patients.

An overview of haemoglobin switching

Haemoglobin is a tetramer composed of both α- and β-like polypeptide subunits. Over the course of ontogeny, the composition of these subunits varies, leading to assembly of haemoglobin molecules with different physiological properties. In humans and old world monkeys, two developmental switches take place for the production of the β-like subunits of the haemoglobin molecule. The initial switch is present in all mammals and involves a switch from haemoglobin subunits expressed exclusively in the transiently-produced embryonic primitive wave of erythrocytes to the haemoglobin subunits produced in the earliest definitive wave of erythrocytes arising from the fetal liver (McGrath & Palis, 2008). This switch is known as the primitive to definitive haemoglobin switch at the β-globin locus. Definitive haemoglobin subunits can be expressed in primitive erythrocytes at low levels, but expression of the primitive embryonic subunits appears to be lineage-restricted (Trimborn et al, 1999; Kingsley et al, 2006; Fraser et al, 2007). It is interesting to note that at the α-globin locus in mammals a similar switch from an embryonic haemoglobin, which is normally restricted to primitive erythrocytes, to the adult α-globin subunits occurs (McGrath & Palis, 2008). However, this switch appears to occur earlier within the primitive lineage (Peschle et al, 1985; Trimborn et al, 1999; Kingsley et al, 2006). Additionally, the lineage restriction of this primitive haemoglobin can be lost in certain pathological conditions (Chui et al, 1986, 1989). This haemoglobin switch will not be discussed further in this review, which is focused on haemoglobin switching at the β-globin locus.

In the majority of mammals that have been well studied, such as mice, the primitive to definitive haemoglobin switch appears to be the predominant event at the β-globin loci (Figs 1 and 2). Occasionally, non-primates have been noted to have evolved additional haemoglobin switches. For example, certain ruminants display additional unique stages of haemoglobin ontogeny (Nienhuis et al, 1974). Sheep and goats have a unique haemoglobin that is normally produced in the late stages of gestation and the early newborn period, which is also induced by anaemia (Huisman, 1974). However, expression of similar stage-restricted haemoglobin subunits is not characteristic of other mammals and it is likely that the molecular mechanisms mediating these switches are unique to this group of mammals. Additionally, the haemoglobin expression pattern in other groups of animals, such as fish and chickens, is often quite different and is not immediately reconciled with human haemoglobin expression (Groudine et al, 1981; Brownlie et al, 2003). These haemoglobin switches will not be discussed further in this review.

Figure 1.

 A diagram illustrating the developmental switching of the β-like globin gene expression in human (left) and mouse (right). Organization of human and murine β-globin loci, consisting of the linked β-like globin genes (coloured boxes), upstream DNaseI hypersensitive sites (HS, red boxes) within the locus control regions (LCR), and downstream 3′HS1, is displayed. Above the graph for the human locus the shifting sites of haematopoiesis are indicated. In the graph for the mouse locus, the content of both endogenous mouse (black straight lines) and exogenous human β-like globins (blue dashed lines) in transgenic β-globin locus mice are shown. This graph is adapted from (Noordermeer & de Laat, 2008).

Figure 2.

 A schematic demonstrating the ontogeny of primitive and definitive erythroid cells from the earliest stem or progenitor cells that give rise to these lineages to more differentiated erythroid progenitors (Ery. Prog.) that then undergo maturation to give rise to mature erythrocytes (McGrath & Palis, 2008; Sankaran et al, 2009). This scheme is shown in panel A. The various globin genes in mouse and humans that are expressed at each corresponding stage are shown below in panel B.

In the course of evolution, old world monkeys acquired a unique stage of haemoglobin expression, reflected by a subunit expressed primarily in the early fetal definitive erythrocytes and then throughout gestation (Johnson et al, 2000, 2002a). In humans and the majority of primates, this fetal haemoglobin subunit is produced by the γ-globin genes (Table I). Some expression of the fetal haemoglobin genes is seen early in ontogeny in the primitive erythrocyte lineage (Peschle et al, 1985). However, with the onset of definitive erythropoiesis from the fetal liver, fetal haemoglobin production is markedly increased (Peschle et al, 1985; Ley et al, 1989). Over the course of gestation, the major β-like haemoglobin subunit that is expressed is γ-globin. γ-Globin is assembled with the adult α-globin subunits to form the fetal haemoglobin (HbF) tetramer (α2γ2). During primate evolution, the genes encoding the fetal haemoglobin subunit were duplicated, such that there are two fetal globin genes in humans, HBG1 (Aγ) and HBG2 (Gγ), which differ by only a single amino acid. As the newborn period approaches, the fetal switch begins to take place from HBG1 and HBG2 (γ-globin) genes to the adult HBB (β-globin) gene (Stamatoyannopoulos, 2005). This switch is normally completed during infancy and typically lasts until approximately 6 months of age (Fig 1). Non-anaemic adults continue to express a low level of HbF, which is largely concentrated in a small percentage of erythrocytes referred to as F cells (Boyer et al, 1975; Thein & Menzel, 2009). Occasionally, in the context of certain pathological conditions or rare mutations, the level of HbF can be elevated. The nature of this variation and the genetics underlying part of this variation has been discussed in a recent review in this journal (Thein & Menzel, 2009). The nature of this switch will be discussed in further detail later in this review.

Table I.   Summary of the Human Gene Nomenclature (HUGO) or Mouse Genome Informatics (MGI) nomenclature and the corresponding conventional gene symbols for the human and mouse β-like globin genes.
Conventional gene symbolsHUGO or MGI nomenclature

Haemoglobin switching of both the primitive to definitive and the fetal to adult types has been studied as models for the developmental control of gene expression. The β-globin loci of mammals were among the first gene loci to be cloned and sequenced, and have constituted an important model system for the study of gene regulatory processes (Fritsch et al, 1980; Leder et al, 1980). Following the cloning of these genes, many advances were made in understanding how gene expression from the β-globin loci is controlled. The function of a powerful upstream enhancer of the β-globin loci, the locus control region (LCR), was identified as essential for high level expression of these genes (Tuan et al, 1985; Forrester et al, 1986; Grosveld et al, 1987; Bender et al, 2000) (Fig 1). Over the ensuing years, many regulators of this gene locus were identified (Cantor & Orkin, 2002). However, the control of the developmental haemoglobin switches remained an enigma.

The importance of understanding the fetal to adult switch in humans is underscored by the clinical relevance to the β-haemoglobin disorders (Stamatoyannopoulos, 2005; Bank, 2006). Infants with sickle cell disease were postulated to be protected from symptoms until several months of age because of elevated HbF levels (Watson, 1948). This notion was substantiated by observations of patients with compound heterozygosity for sickle cell disease and hereditary persistence of fetal haemoglobin (HPFH) mutations who were largely asymptomatic (Weatherall & Clegg, 2001). Related observations were made in patients with β-thalassaemia mutations, where higher levels of HbF correlate with a more asymptomatic clinical course (Weatherall, 2001). These observations in patients with disorders of the β-haemoglobin subunit have been further supported in numerous diverse populations and in larger epidemiological studies (Perrine et al, 1972; Platt et al, 1991, 1994; Weatherall & Clegg, 2001; Premawardhena et al, 2005).

The principle that elevated HbF ameliorates the severity of the β-haemoglobin disorders has been the driving force behind efforts to stimulate fetal haemoglobin production in humans. In the early 1980s, initial trials of 5-azacytidine were performed and this subsequently led to the trials using hydroxycarbamide (which was tested because it appeared to have less cytotoxic side effects compared with 5-azacytidine) for the induction of HbF in patients with haemoglobin disorders (Ley et al, 1982, 1983; Platt et al, 1984; Platt, 2008). While hydroxycarbamide has been used successfully in patients with sickle cell disease, it is not uniformly effective and rarely ameliorates all symptoms of the disease (Platt, 2008). It should also be noted that the clinical efficacy of hydroxycarbamide may not only relate to its ability to increase HbF levels, but may be due to other effects it has on red blood cells, as well as white blood cells in the circulation (Platt, 2008). There may additionally be some benefit to the use of hydroxycarbamide as an HbF inducer in select patients with β-thalassaemia (Yavarian et al, 2004). A variety of other agents have also been used clinically to induce HbF and this has been detailed in a recent review in this journal (Trompeter & Roberts, 2009). However, neither uniformly effective therapies, nor any targeted therapies aimed at reversing this fetal to adult haemoglobin switch have yet been developed, highlighting the importance of gaining a greater understanding of the control of the fetal to adult haemoglobin switch in humans.

The embryonic primitive to definitive haemoglobin switch

As described above, all mammals have two distinct erythroid lineages. The primitive erythroid lineage originates in the yolk sac and undergoes terminal maturation in the bloodstream, fetal liver and other reticuloendothelial organs (Fig 2) (McGrath & Palis, 2008). This transient population is then replaced by the definitive erythroid lineage that generates smaller enucleated erythrocytes that play the predominant role in oxygen transport throughout gestation and postnatal life (McGrath & Palis, 2008). The definitive erythroid lineage is continuously produced by hematopoietic stem cells and their more differentiated progeny (Orkin & Zon, 2008). A distinguishing feature of the primitive and definitive lineages is the presence of different globin gene expression patterns. This was first noted in studies of globin gene regulation during early embryonic development in humans and mice (Peschle et al, 1985; Whitelaw et al, 1990). A set of primitive embryonic globin genes is solely expressed within the primitive erythrocytes (Trimborn et al, 1999; Kingsley et al, 2006; McGrath & Palis, 2008). As primitive cells are progressively replaced in fetal development by the definitive erythroid cells, a switch from the embryonic globin genes to the definitive globin genes occurs (Figs 1 and 2). Of note, there is some expression of the definitive erythroid globin genes within the primitive erythroid lineage, but this expression is at low levels and the extent of expression increases dramatically with robust production of red cells from the definitive lineage (Kingsley et al, 2006). Additionally, a process termed ‘maturational’ globin switching appears to occur within this lineage, at least in the context of mice that have multiple embryonic primitive globin genes. In this species, the initially highly-expressed Hbb-bh1 (βh1) embryonic globin is superseded within this lineage by the Hbb-y (εy) globin during the maturation process of these cells (Kingsley et al, 2006) (Fig 1). A similar maturational phenomenon has also been noted to occur at the α-globin locus (McGrath & Palis, 2008).

The HbF switch

The fetal to adult switch that occurs in humans and old world monkeys is unique. The rationale for the evolution of this switch is unclear, although the presence of HbF has been thought to be necessary to facilitate transplacental oxygen exchange. Of note however, is the fact that placental mammals with gestation times that are greater than humans do not always have a unique fetal-type haemoglobin. However, the presence of an HbF expression stage has provided a natural factor for phenotypic variation among humans and a target for therapies in the β-haemoglobin disorders. This switch was noted to take place in humans and old world monkeys after birth (Johnson et al, 2000). As a result, intense efforts have been aimed at understanding the molecular mechanisms that underlie this developmental switch.

In the early studies of the switching mechanism, debate centred on the nature of the switch from fetal to adult haemoglobin. It was argued that this switch appeared to be consistent with a change in cell lineages, such that a group of cells that initially expressed only fetal haemoglobin was progressively replaced by separate lineages that expressed either a mixture of fetal and adult haemoglobin or eventually entirely adult haemoglobin (Alter et al, 1983). However, subsequent evidence argued that the same progenitors give rise to all of the progeny with variations in expression of fetal and adult haemoglobin (Stamatoyannopoulos, 2005). This work involved a variety of approaches, including clonal erythroid cell cultures and colony assays, chromosomal hybrids, transplantation assays and analysis of clonal hematopoietic disorders. It appears, therefore, that the switch from fetal to adult haemoglobin occurs in a clonal population rather than through replacement by different lineages (Fig 2).

Subsequently, a variety of models for the molecular control of switching were postulated. Early work in the 1960s suggested models in which haemoglobin switching could occur either by maturational replacement of globin genes from different loci or an actual genetic switch where various genes at single loci could be turned on and off (Baglioni et al, 1961; Zipursky, 1965). With the cloning of the haemoglobin genes, it became evident that the latter models were valid (Fritsch et al, 1980). In the ‘molecular’ era models have invoked gene silencing, gene competition, chromosomal looping and tracking to account for the fetal to adult globin switch (Orkin, 1990; Stamatoyannopoulos, 2005; Bank, 2006). While the literature exists to support different models, it is important to recognize that the various models are not necessarily mutually exclusive. Moreover, it is likely that each model oversimplifies what must occur in an actual erythroid progenitor during globin gene expression. For example, while strong evidence supports looping of the upstream LCR to the downstream genes within the β-globin locus, looping by itself does not preclude transient tracking within the locus or autonomous gene silencing of individual globin genes (Tolhuis et al, 2002; Palstra et al, 2003; Vakoc et al, 2005). Additionally, it is important to recognize that experimental evidence for specific models has largely relied on the use of mice harbouring human globin transgenes in various configurations. Transgenic mice, though generally useful, may not faithfully mimic the human developmental context. Recently presented models of gene regulation underscore how these early models may not reflect the complexity of transcriptional regulation occurring in the intact cell (Lieberman-Aiden et al, 2009). While this review is focused primarily on the developmental transcriptional switch in haemoglobin, classical studies suggest that additional, as yet unknown, mechanisms may constitute a developmental ‘clock’ that has been suggested to function at the cis-acting level (Wood et al, 1985; Papayannopoulou et al, 1986). Such a mechanism may pertain more to silencing of HbF expression in adults rather than during the switching process in development.

Natural variation in the level of HbF among individuals has been long recognized, and appears to reflect differential levels of transcription of the fetal haemoglobin genes (Stamatoyannopoulos, 2005; Thein & Menzel, 2009). Moreover, in conditions such as stress erythropoiesis, which is thought to involve alterations in cell cycle regulation and maturation of erythroid progenitors, HbF levels are also elevated (Alter et al, 1976; Papayannopoulou et al, 1980). These latter observations contributed to therapeutic efforts to modulate the cell cycle of erythroid cells as a means to therapeutically reactivate HbF in the β-haemoglobinopathies. This path of investigation led to the discovery of hydroxycarbamide as an inducer of HbF in patients with sickle cell disease (Papayannopoulou et al, 1984; Letvin et al, 1985; Stamatoyannopoulos, 2005). The precise connection between altered cell cycle kinetics of erythroid cells and altered globin gene expression is still not well understood at the molecular level (Stamatoyannopoulos, 2005; Higgs & Wood, 2008; Sankaran et al, 2008a). Further work on the effectors of this stress erythropoietic response of HbF, which appears to unique in humans and old world monkeys (Sankaran et al, 2009), may lead to important insight into haemoglobin switching.

More recently, recognized variability in the expression of HbF both in non-anaemic individuals and patients with β-haemoglobin disorders stimulated efforts to utilize new human genetic approaches to delineate factors that might explain this variation (Thein & Menzel, 2009; Thein et al, 2009). These efforts have focused on genetic association studies. In general, genetic association studies use unrelated individuals to ascertain correlations between a specific phenotype and variation at common genetic polymorphisms (with minor allele frequencies >5%) (Thein & Menzel, 2009). The first gene locus unlinked to the β-globin locus that regulates HbF levels in humans was found through the use of genetic association analyses in a candidate region initially identified through Mendelian genetic approaches (Thein et al, 2007; Thein & Menzel, 2009). As a result, common variants situated between the HBS1L and MYB genes were found (Thein et al, 2007; Lettre et al, 2008; So et al, 2008; Uda et al, 2008). The biological effect of this genetic variant is uncertain, although studies suggest that alterations in the level of the MYB product may be responsible for the elevation in HbF seen with these variants (Jiang et al, 2006; Wahlberg et al, 2009). Soon thereafter, two genome-wide association studies (GWAS) studies sought common genetic variants affecting HbF in independent non-anaemic populations (Menzel et al, 2007; Uda et al, 2008). These studies confirmed the effect of the HBS1L-MYB variants and also identified a new set of variants in an intron of the gene BCL11A. Subsequently, these effects were confirmed in populations with sickle cell disease (Lettre et al, 2008) and other populations (Sedgewick et al, 2008). Additionally, the effect of these HbF-regulating variants on clinical severity has been shown for both sickle cell disease and β-thalassaemia (Lettre et al, 2008; Uda et al, 2008; Galanello et al, 2009; Nuinoon et al, 2009). These findings led to the work demonstrating that BCL11A is a developmental stage-specific regulator of the fetal to adult haemoglobin switch in humans, as will be discussed below (Sankaran et al, 2008b).

Role of molecular factors involved in the haemoglobin switches

A variety of nuclear factors involved in transcriptional regulation have been suggested to be involved in globin gene regulation and switching. In this section of this review we discuss the evidence relating these factors to haemoglobin switching:


The zinc-finger transcriptional factor BCL11A (B-cell lymphoma/leukaemia 11A, also known as EVI9, CTIP1) was initially cloned as a myeloid or B cell proto-oncogene in mice and humans, respectively (Fell et al, 1986; Li et al, 1999; Nakamura et al, 2000; Suzuki et al, 2002). Bcl11a mutant embryos lack B cells and have alterations in several types of T cells, indicating that BCL11A is indispensible for normal lymphoid development (Liu et al, 2003). A potential role for BCL11A in the red blood cell lineage, particularly in haemoglobin switching, was first suggested by genetic association studies of HbF levels in humans (Menzel et al, 2007; Lettre et al, 2008; Uda et al, 2008). Using a variety of approaches, BCL11A has been validated as a major regulator of HbF switching and silencing in humans (Sankaran et al, 2008b). Of note, BCL11A exhibits stage-specific expression in human ontogeny, such that fetal and embryonic cells that robustly express HbF, express shorter variant forms of this protein. The genetic variant within the BCL11A locus, maximally associated with higher levels of HbF, is associated with reduced expression of BCL11A mRNA. In erythroid progenitors, BCL11A physically interacts with the NuRD chromatin remodelling complex, and the erythroid transcription factors, GATA1 and FOG1. Knockdown of BCL11A in cultured human erythroid progenitors leads to robust HbF expression, consistent with a role of BCL11A as a repressor of HBG1 and HBG2 and as a key regulator of the fetal to adult haemoglobin switch in humans (Sankaran et al, 2008b). On knockdown of BCL11A, few expression differences are seen between knockdown and control cells, suggesting that BCL11A acts rather specifically and perhaps directly at the β-globin locus. BCL11A occupancy of adult erythroid chromatin in several discrete sites within the β-globin locus, including regions within the LCR and downstream of HBG1 in the HBG1-HBD-intergenic region, supports a direct role. Taken together, these findings establish BCL11A as repressor of HbF expression and critical for the maintenance of HbF silencing in adult human erythroid cells.

More recent work in the mouse provides persuasive evidence for a central role for BCL11A in the developmental switching of haemoglobin genes (Sankaran et al, 2009). As noted above, the mouse exhibits a single embryonic-to-adult switch and lacks a fetal-gene equivalent. Therefore, its switching pattern is not strictly analogous to that of human. Early work suggested that a HBG1-HBG2 transgene (lacking the LCR) was expressed at exceedingly low levels but in pattern similar to that of mouse embryonic β-like genes (Chada et al, 1986). Reevaluation of transgenic mice containing the entire human β-globin cluster as a yeast artificial chromosome extended this observation by showing that high level HBG1 and HBG2 expression is present primarily in the primitive lineage and extinguished very early in the fetal liver stage, unlike in the human (Fig 1). While these findings point to important limitations to the transgenic model of globin gene regulation, they also argue that critical alterations in the trans-acting environment have occurred in the course of evolution. Of relevance to this conclusion, expression of BCL11A also differs between mouse and human, suggesting that BCL11A may constitute an important evolutionary driver of divergent switching. In mice, BCL11A expression is restricted to the definitive erythroid lineage, whereas in human it is expressed in the pattern discussed above with expression of full length isoforms only being noted in postnatal definitive erythroid cells. To interrogate potential roles for BCL11A in developmental switching of globins, transgenic mice harbouring the human β-globin cluster were interbred with mice carrying a Bcl11a knockout allele. This strategy permits assessment of both endogenous mouse and exogenous human β-like globin expression in the absence of BCL11A. In Bcl11a knockout mice, restriction of expression of the mouse embryonic Hbb-y and Hbb-bh1 genes to the primitive lineage is lost. Additionally, in mice with the human β-globin locus transgene (as a yeast artificial chromosome clone, β-YAC), robust HBG1 and HBG2 expression occurs within the definitive erythroid lineage. Based on these observations, it appears that BCL11A restricts β-like embryonic globin expression to the primitive lineage in the mouse, whereas in the human it restricts fetal haemoglobin expression to the fetal and newborn period. Together these findings suggest that BCL11A or its partner proteins may serve as excellent leads for targeted therapeutic approaches to reactivate HbF in patients with β-haemoglobinopathies. Downregulation of BCL11A expression or impairment of BCL11A function could be promising strategies.


SOX6 belongs to the family of Sry-related HMG box transcription factors, many of which are best known as determinants of cell fate and differentiation in various lineages (Wegner, 1999; Schepers et al, 2002). SOX6 is expressed in several tissues, including cartilage, testis, neuronal and erythropoietic tissues (Connor et al, 1995; Takamatsu et al, 1995; Lefebvre et al, 1998; Hagiwara et al, 2000; Dumitriu et al, 2006). Mice featuring a chromosomal inversion (p100H) resulting in inactivation of Sox6 gene, and mice with a targeted inactivation of Sox6 die as neonates secondary to cardiac or skeletal myopathy (Hagiwara et al, 2000; Smits et al, 2001). A potential role of SOX6 in haemoglobin gene regulation was first recognized by analysis of the Sox6-deficient mouse. At the fetal liver stage, both mouse embryonic β-like globins are dramatically elevated in the Sox6-deficient p100H mouse (Yi et al, 2006). While Hbb-bh1 was downregulated rapidly in late fetal livers, Hbb-y was persistently expressed until around birth (Yi et al, 2006). SOX6 also represses Hbb-y and, to a lesser extent, Hbb-bh1 expression in definitive erythropoiesis of adult mice, as shown by transplanting fetal liver cells from Sox6-deficient mice into wild-type adult mice (Cohen-Barak et al, 2007). SOX6 may act as repressor of Hbb-y expression by directly binding to the Hbb-y promoter (Yi et al, 2006). Intriguingly, SOX6 may also play a role in erythroid cell maturation as increased numbers of nucleated red cells are present in the fetal circulation of Sox6-deficient embryos (Yi et al, 2006). SOX6 has also been suggested to enhance definitive erythropoiesis in mouse by stimulating erythroid cell survival, proliferation and terminal maturation (Dumitriu et al, 2006). While the exact role of SOX6 in haemoglobin switching remains to be determined, SOX6 has been reported to be able to act as either an activator or a repressor, depending on its interacting proteins and promoter context (Lefebvre et al, 1998; Murakami et al, 2001). In addition, Sox transcription factors bind to the minor groove of DNA and cause a drastic bend of the DNA that leads to local conformational changes (Ferrari et al, 1992; Connor et al, 1994). Therefore, SOX6 may perform part of its function as architectural protein by organizing local chromatin structure and associated DNA-bound transcription factors into biologically active, sterically defined multiprotein complexes (Yi et al, 2006). The role of SOX6 in human globin gene regulation has not yet been examined directly, although recent work in human erythroid progenitors suggests that variation in the levels of SOX6 may play a role in repressing HbF expression (Sripichai et al, 2009).


The GATA family of proteins (GATA1–6) comprises zinc-finger transcription factors that both activate and repress target genes containing a consensus GATA binding motif (Orkin, 1992). Binding sites with this motif are present in many positions in the α- and β-globin loci, as well as many other erythroid-expressed genes. The founding member of this family, GATA1, was discovered as a β-globin locus-binding protein (Evans & Felsenfeld, 1989; Martin et al, 1989; Tsai et al, 1989). GATA1 is essential for erythroid cell maturation in vivo (Pevny et al, 1995). Gata1-null cells of either the primitive or definitive lineage fail to mature beyond the proerythroblast stage (Pevny et al, 1995; Weiss & Orkin, 1995; Blobel & Orkin, 1996; Fujiwara et al, 1996). GATA1 has been suggested to activate expression of the adult mouse Hbb-b1 by recruiting RNA polymerase II to the Hbb-b1 promoter (Johnson et al, 2002b). GATA1 appears to facilitate chromatin loop formation at both the Kit and β-globin loci (Vakoc et al, 2005; Jing et al, 2008). Thus, GATA1 might participate in haemoglobin switching by inducing chromatin looping. In support of this hypothesis, GATA1 has been shown to bind a region upstream of both the HBG1- and HBG2-promoter in a FOG1 dependent manner, leading to recruitment of the repressive NuRD-complex (Harju-Baker et al, 2008). This region upstream of the HBG1 and HBG2 promoters has been suggested to be necessary for HbF silencing in transgenic mouse models (Harju-Baker et al, 2008). A direct role for GATA1 in regulation of the fetal to adult haemoglobin switch in humans has been suggested by the report of a patient with congenital erythropoietic porphyria and elevated HbF who had a GATA1 zinc-finger mutation (Phillips et al, 2007). However, the precise aetiology for elevated HbF in this patient is not clear.


Kruppel-like factor 1 [KLF1 (erythroid), also known as EKLF] was discovered as an activator of adult HBB through a highly conserved CACCC motif that was known to be mutated in human β-thalassaemias (Miller & Bieker, 1993). Klf1-null mice appeared to be embryonic-lethal due to a marked reduction in Hbb expression, while embryonic globin genes were expressed normally (Nuez et al, 1995; Perkins et al, 1995). Therefore, KLF1 was initially thought to be an adult stage-specific factor that facilitates human fetal to adult haemoglobin switching (Donze et al, 1995; Perkins et al, 1996; Wijgerde et al, 1996; Gillemans et al, 1998). However, subsequent studies in Klf1-null transgenic mice indicate that KLF1 is also active during the primitive stage of erythropoiesis. An LCR-β-globin transgene, which is normally expressed in primitive erythroid cells, is not expressed in the absence of KLF1, indicating that KLF1 is also a transcriptional activator in primitive erythropoiesis (Guy et al, 1998; Tewari et al, 1998). Therefore, it remains to be determined whether KLF1 is directly involved in haemoglobin switching. Besides regulating β-globin expression, studies on Klf1-null mice demonstrated the crucial role of KLF1 in regulating both definitive and primitive erythropoiesis (Nuez et al, 1995; Perkins et al, 1995; Hodge et al, 2006).

Kruppel-like factor 1 occupies HS1-HS3 of the human β-globin LCR and the Hbb-b1 promoter in mouse erythroid cell lines (Im et al, 2005). In mouse primitive and definitive erythroid cells from E10.5 yolk sac, E14.5 fetal liver and Ter119+ bone marrow cells, KLF1 also occupies HS1-HS4 and β-like globin promoters in a differentiation stage-specific manner (Zhou et al, 2006). KLF1 interacts physically and functionally with CBP/p300 (Zhang et al, 2001), and BRG1 (Armstrong et al, 1998; Tewari et al, 1998; Brown et al, 2002), a component of the SWI/SNF chromatin remodelling complex. However, BRG1 distribution at the β-globin locus does not correlate precisely with the KLF1 occupancy pattern, indicating that KLF1 is not the sole determinant of BRG1 occupancy at the locus (Im et al, 2005; Wozniak et al, 2008). CBP/p300 binds and acetylates KLF1 and thereby may regulate its transactivation activity (Zhang et al, 2001). Although these results remain to be confirmed in vivo, the acetylation status of KLF1 may determine its effects at different stages of ontogeny (Chen & Bieker, 2004). Targeted disruption of Klf1abrogates DNaseI hypersensitivity at HS3 and HBB promoter (Tewari et al, 1998), indicating that KLF1 regulates chromatin structure at these sites. Consistent with this notion, it has been shown that KLF1 promotes the formation of a high-order chromatin loop at the β-globin locus (Drissen et al, 2004). In addition, KLF1 also functions as a transcriptional repressor by recruiting Sin3A and HDAC1 (Chen & Bieker, 2001, 2004). Whether these effects of KLF1 on haemoglobin gene expression are directly mediated at the locus or are due to indirect effects on erythroid maturation remains to be determined in future studies. Additionally, the role of KLF1 as a regulator of haemoglobin switching in human cells has yet to be assessed directly. It is interesting to note that a recent report of a patient with features of congenital dyserythropoietic anaemia and elevated HbF levels at c. 40% appears to have a mutation in KLF1, which may contribute to this phenotype, suggesting a role of this factor in human fetal haemoglobin switching (Singleton et al, 2009). Further study of unique patients of this sort will probably provide critical insight into the mechanisms of haemoglobin switching in humans.


NF-E4 was first described as a critical transcription factor in chicken globin switching (Choi & Engel, 1988; Gallarda et al, 1989; Yang & Engel, 1994). A human homologue of chicken NF-E4, p22NF-E4, has been suggested to be active in human fetal globin gene activation (Zhao et al, 2004; Zhou et al, 2004). p22NF-E4, together with a ubiquitous transcription factor CP2, are part of a fetal erythroid transcription factor complex known as the stage selector protein (SSP) (Jane et al, 1995; Zhou et al, 2000). Ectopic expression of p22NF-E4 in K562 cells and human cord blood progenitors increases HBG1 and HBG2 expression. Enforced expression of p22NF-E4 in human β-locus transgenic mice delays the fetal to adult switch (Zhou et al, 2004). However, the switch in globin subtype is fully completed in the adult bone marrow in these mice. These findings suggest that p22NF-E4 is capable of influencing human globin gene expression in vivo in mouse models, but is incapable of overriding the intrinsic mechanisms governing γ-globin silencing in this context. Currently the role of p22NF-E4 in regulating human haemoglobin switching is unclear and further work investigating this factor is needed to establish its relevance.


The proximal promoters of the HBE1- and HBG1/HBG2, but not HBB contain direct repeat sequences analogous to DR1 binding sites for non-steroid nuclear hormone receptors. The DR1 sites reside near the CCAAT-box, one of the most conserved motifs found in globin promoters, and a region affected by the -114 and -117 HPFH mutations (Filipe et al, 1999). Mutations of the DR1 sites that inhibit protein binding result in HBE1 expression in the adult transgenic mice, indicating that the DR elements and the associated DNA-bound factors participate in HBE1 silencing (Filipe et al, 1999; Tanimoto et al, 2000). DR elements bind a factor immunologically related to COUP-TF orphan nuclear receptors. One of these, COUP-TFII (also called ARP-1), is expressed in embryonic/fetal erythroid cell lines, mouse yolk sac and fetal liver (Filipe et al, 1999). More recently, it was suggested that COUP-TFII functions as a downstream repressor of HBG1 and HBG2 after stem cell factor (SCF) stimulation in cultured human adult erythroid progenitors (Aerbajinai et al, 2009). In these cells, expression of COUP-TFII is suppressed by SCF through phosphorylation of serine/threonine phosphatase (PP2A), and downregulation of endogenous COUP-TFII expression leads to increase in HBG1 and HBG2 expression (Aerbajinai et al, 2009). Therefore, SCF signalling may activate HBG1 and HBG2 expression by inhibition or removal of the repressor COUP-TFII that binds to the DR elements at the HBG1 and HBG2 promoters. Interestingly, BCL11A was also isolated as an interacting partner (CTIP1) of the orphan nuclear receptor COUP-TFII (Avram et al, 2000). It remains to be determined whether BCL11A and COUP-TF cooperate in silencing HBG1 and HBG2expression during erythroid development.


The DRED (direct repeat erythroid-definitive) complex was initially identified as a definitive-stage HBE1 repressor with strong affinity to DR1 sites in the human HBE1 promoter (Tanimoto et al, 2000). The binding of DRED complex to the DR elements in definitive erythropoiesis is postulated to prevent KLF1 from activating HBE1. DRED contains the nuclear orphan receptors TR2 and TR4, which form a heterodimer when they bind to the DR1 sites in the human HBE1 and HBG1/HBG2 promoters (Tanimoto et al, 2000; Tanabe et al, 2002). The human HBB gene lacks DR1 sites, suggesting that repression of HBE1 and HBG1/HBG2 expression may be the major function of DRED complexes in adult cells. In support of this hypothesis, a point mutation at -117 of the HBG1 promoter (HPFH) associated with high HbF disrupts one DR1 site and reduces TR2/TR4 binding (Tanabe et al, 2002). TR2 and TR4 also bind to an evolutionally conserved DR element within the GATA1 hematopoietic enhancer (G1HE) and directly repress GATA1 transcription, suggesting a mechanism by which GATA1 may be directly silenced by TR2/TR4 during terminal erythroid maturation (Tanabe et al, 2007). The role of this complex in regulating the fetal to adult -globin switch in human cells has not yet been directly assessed and future work will be necessary to determine its contribution to this process.


MBD2 belongs to a group of proteins characterized by the conserved methyl-CpG binding domain (MBD), which is necessary and sufficient for binding of the protein to methylated DNA (Nan et al, 1993). MBD2 is part of the methyl-CpG binding protein complex 1 (MeCP1), which contains the proteins Mi-2, MTA1, MTA2, MBD3, HDAC1, HDAC2, RbAp46 and RbAp48 (Feng & Zhang, 2001). MBD2 and DNA methylation were suggested to be involved in the normal developmental regulation of chicken embryonic ρ-globin gene (Ginder et al, 1984; Burns et al, 1988; Singal et al, 1997, 2002). The ρ-gene is highly methylated and enriched for MBD2 in adult erythrocytes when the gene is silent, but not when it is actively transcribed. Studies in Mbd2-null mice with a transgenic human β-locus (β-YAC) indicate that DNA methylation and MBD2 are required to repress HBG1 and HBG2 in adult erythroid cells. Adult Mbd2−/−/β-YAC mice displayed elevated HBG1 and HBG2 expression at a level commensurate with 5-azacytidine treatment (Rupon et al, 2006). The level of HBG1 and HBG2 expression is consistently higher in Mbd2-null mice at E14.5 and E16.5 fetal liver erythroblasts, indicating a delay in fetal haemoglobin silencing during embryonic/fetal erythroid development. Of note, however the induction seen in this model is much lower (c. 20-fold compared with >1000-fold) than what was recently shown in the context of knocking out BCL11A in similar transgenic mouse models (Sankaran et al, 2009). In addition, MBD2 does not appear to bind directly to the HBG1 and HBG2 promoter region while DNA methylation levels are modestly decreased in Mbd2-null mice (Rupon et al, 2006).

Ikaros-PYR complex

Ikaros was first implicated as a factor involved in haemoglobin switching through its presence in a chromatin remodelling complex (called PYR complex) that is specifically present in adult mouse and human hematopoietic cells (O’Neill et al, 1999, 2000; Bank, 2006). PYR complexes consist of the SWI/SNF and NuRD chromatin remodelling complexes and bind to a 250-bp polypyrimidine-rich DNA sequence upstream of the human HBD gene. The apparent PYR complex DNA-binding site is included in the Corfu deletion (O’Neill et al, 1991). Mice that lack Ikzf1 (Ikaros; Ik−/−) have no PYR complex binding activity, indicating the requirement for Ikaros in the formation of the complex on DNA (Lopez et al, 2002; Keys et al, 2008; Bottardi et al, 2009). Ikzf1-null mice exhibit a modest (1–2 d) delay in mouse embryonic to adult β-like haemoglobin switching, and in human fetal to adult globin switching in Ikzf1-null mice crossed with transgenic mice containing the human β-globin locus (Lopez et al, 2002; Keys et al, 2008). However, the null mice also exhibit multiple hematopoietic cell defects including anaemia and megakaryocytic abnormalities (Lopez et al, 2002). Whether the delayed haemoglobin switching seen in the transgenic mice is due to a direct effect or due to an indirect effect on erythropoiesis is uncertain.


ATP-dependent chromatin remodelling complexes, such as the SWI/SNF complex, serve important roles in transcriptional activation (Saha et al, 2006). SWI/SNF uses ATP hydrolysis to disrupt histone-DNA interactions, therefore increasing factor accessibility to nucleosomal DNA (Narlikar et al, 2002). The BRG1 catalytic subunit of the SWI/SNF complex is essential in developmental and physiological processes (Bultman et al, 2005; de la Serna et al, 2006). The SWI/SNF complex has been suggested to function directly to regulate β-like globin gene expression (O’Neill et al, 1999, 2000; Lopez et al, 2002). BRG1 occupies HS2-4 and the Hbb-b1 promoter at the murine β-globin locus, and GATA1 increases BRG1 occupancy at a subset of these sites (Im et al, 2005). Although the precise mechanisms by which the SWI/SNF complex regulates haemoglobin expression remains unknown, BRG1 associates with several factors implicated in HBB regulation. BRG1 binds KLF1, and the SWI/SNF complex is required for KLF1-mediated transcriptional activation in vitro (Armstrong et al, 1998; Kadam et al, 2000). BRG1 is part of the MafK complex in MEL cells and also a β-globin LCR-associated chromatin remodelling complex known as LARC (Brand et al, 2004; Mahajan et al, 2005). In addition, the SWI/SNF complex is also part of the PYR complexes which has been implicated in regulating human fetal to adult globin gene switching in transgenic mouse models (O’Neill et al, 1999, 2000; Bank, 2006). Recent work using hypomorphic mouse alleles of BRG1 suggest that it may have a role in haemoglobin switching, although these mice also are noted to have defects in erythropoiesis (Bultman et al, 2005). As with factors such as GATA-1, SOX6, KLF1 and Ikaros, separating the direct effect of this factor on globin gene regulation versus any indirect effects it may have on erythroid development confounds simple interpretation of existing data.

Concluding remarks

Until recently the molecular mechanisms mediating haemoglobin switching have remained largely elusive. However, the recent discovery and validation of BCL11A as a haemoglobin switching factor should stimulate renewed efforts to dissect the process in detail and search for approaches to induce HbF in a targeted manner. We anticipate that further studies of BCL11A will lead to identification of cooperating factors and a set of additional targets for therapeutic intervention.

Several important lessons may be derived from the recent history of the haemoglobin switching field. First, inputs from seemingly unrelated disciplines may be necessary to overcome obstacles in understanding a problem. The identification of BCL11A as a candidate regulator emerged from the tools of contemporary genome-wide association studies rather than from within more traditional haematological approaches (Michelson, 2008; Sankaran et al, 2008b). Second, tools that are often accepted as the ‘goal standard’ within a field may have unappreciated limitations that hamper a full appreciation of a biological problem. For example, mice containing the intact human β-globin locus as a transgene, which have represented the major workhouse for the field for two decades, are not ideal models due to inherent differences in haemoglobin switching between species (Sankaran et al, 2009). Nonetheless, once potential shortcomings are acknowledged, experimental findings can be informative. Third, in order to study processes unique to old world monkeys and humans, one will probably need to incorporate study of these species to further translational applications of basic findings. Fourth, while we are starting to understand factors that directly regulate the β-globin loci, the molecular connections to phenomena such as stress erythropoiesis and alterations in cell cycle kinetics remain to be established.

While haemoglobin switching is likely to remain an important model system for basic features of developmental gene regulation, we are hopeful that recent findings will stimulate work that will culminate in the development of novel, effective therapies for patients with β-haemoglobinopathies. Strategies including (but not limited to) the use of small molecule inhibitors or gene therapy knockdown approaches with shRNAs aimed at targeting BCL11A (or other regulators of haemoglobin switching) are promising avenues for the field. Over three decades have passed from the cloning of the β-globin loci to our current understanding of haemoglobin switching. We hope that in the future, progress in this field, ranging from the bench to the bedside, may proceed more rapidly for the sake of patients with β-haemoglobin disorders.